Learning Center
Plans & pricing Sign in
Sign Out

RTT min RTT Network Research Lab


									Inline Path Characteristic Estimation
     to Improve TCP Performance
 in High Bandwidth-Delay Networks

  Cesar Marcondes   HIDEyuki Shimonishi
   Anders Persson     Takayuki Hama
   Medy Sanadidi
    Mario Gerla       Tutomu Murase
• Inferring Path Characteristics
   – Bandwidth Estimation
   – Buffer Estimation
   – Description of Testbed
• Improving TCP Performance using Inline Path
   – TCP Westwood BBE (Buffer and Bandwidth Estimation)
• Conclusions and Future Work
        - Inferring Path Characteristics -

    Buffer Estimation           Bandwidth Estimation
    Loss discrimination:             Link capacity:
  Efficiency in leaky pipes     Efficiency in large pipes
    Congestion detection:
  Friendliness to TCP-Reno

Design novel protocols that make wise usage of estimations
Internet Measurement Testbed Topology
           (Transpacific Link)
                  Capacity Estimation
• Pathrate - Active                     • Capprobe - Active
    (Infocom 2001)                         (Sigcomm 2004)
     – dispersion of packet pairs          - Packet pair measurement with
                                           filtering out delayed samples
        and packet trains, uncover
        possible capacity modes,
        apply statistic analysis to

                  TCP Inline Measurements
•TCP-Westwood                  •TCPProbe
 (Global Internet 2005)        (Global Internet 2005)
   - Packet pair like          - TCP inline version of CapProbe, it flips
capacity estimation using      packets to obtain packet pairs in a delayed
TCP ACK packets                ACK scenario 5 3 4 2 1
 Measurement results (1)
       Off-line Measurement

Convergence Speed

    70sec           6sec      12sec
    Measurement results (2)
       TCP Inline Measurement

                Accuracy   Convergence
Original TCPW   ≈ 60Mbps     < 1sec
  TCPProbe      ≈ 80Mbps      32sec
               Buffer Estimation
• Idea: Correlate the loss of a packet with the RTT
  observed just before the loss
   – Buffer size estimation (RTT value a packet loss likely
     to happen)
   – Congestion estimation, loss discrimination (position of
     current RTT)

                                                  Buffer size
             Measurement results (1)
     Off-line measurement for buffer capacity
•Traceroute + Heavy Cross-Traffic Measurements
     track down the location of bottleneck and observed an increase of 15msec
on the average ICMP delay
•RTT histogram of TCP flows
                                             Buffer size = 15msec

                                                              High load >80%
                   Low load <10%                             Aggregate 10 Flows
                    Single Flow

Overall loss rate = 0.005%                Loss rate (RTT<147msec)
                                          = 0.005%
                      Random loss ?
              Measurement results (2)
     Off-line measurement for random losses
•Simulation Results
   •135msec RTT + 15msec buffer
   •80Mbps bottleneck like
   •4.9 * 10-5 uniform dist. random loss

           1 flow                                        10 aggregated flows
           Simulation 17Mbps                             Simulation 72Mbps
           Measurement 15Mbps                            Measurement 65Mbps

•Low Rate UDP Measurement
   • 1-3Mbps UDP CBR traffic               0.001-0.005% should be a good number
    Loss rate = 0.001%                    for random loss rate
          TCP Westwood BBE
• Optimize tradeoff between efficiency and
  friendliness to TCP-Reno
  – Robustness to buffer capacity variations and
    RTT variations

• Optimize window reduction upon a packet
  loss, using
  – In line bandwidth estimation
  – In line buffer estimation
    TCPW-BBE Buffer Estimation
• RTT dynamic range
   – RTTmin: minimum RTT (= propagation delay)
   – RTTcong: RTT value when packet losses likely to happen
• Congestion = position of current RTT in the range
      RTTmin                         RTTcong
                                                   RTTcong is an exponential average
                                                   of RTT right before losses
                                                    Congestion level
               Losses due to congestion                  RTT – RTTmin
                                                        RTTcong – RTTmin
               Losses due to error
     RTT before packet losses [msec]
         Congestion Window Reduction
Congestion window reduction
    after a packet loss
                                           e c LinkBandwidth  (1  e c ) SendingRate
                                                   1 EligibleRateEstimate
                                                  1 c   SendingRate
               TCP-Reno                                           Original TCPW

           0                           1             (RTT=2xRTTmin)

  Should be a random loss     Should be a congestion loss
     Smaller reduction            Halve the window
     Measurement Result (1)
Buffer estimation and loss discrimination
                     • Buffer Estimation
                        – Inline estimating of buffer
                          around 15 msec
                        – * consistent with previous

                     • Inline Loss Discrimination
                        – Window reduction
                          depending on the
                          congestion level estimated
                        – * more robust to non-
                          congestion errors
                    Measurement Result (2)
                       Efficiency and Friendliness

Average Throughput of Multiple Trials    Cumulative Throughput during 2000 Secs
Pathload Estimation: 61-74Mbps Avail.    Pathload Estimation : 30Mbps Avail.
Conjectured Non-Congestion Error: 10-5   More Congestion losses
     Conclusions and Future Work
•   Path Characteristic Estimations as presented in this paper are reliable/fast/non-
    intrusive and present a major advantage for future Intelligent Network Stacks

•   In our measurement study, we presented, in addition to extensive Internet
    results, cross-validation based on more than one source and we observed that
    the inline capacity and buffer estimations are reliable.

•   In the specific scenario tested, TCP BBE was able to improve the performance
    when comparing to NewReno on a non-congested scenario

•   In the future, a combination of BE (Bandwidth Estimation from TCP
    Westwood) and TCPProbe can lead to faster capacity estimation. Additionally,
    new methods of using such estimates to improve start-up, cross-traffic
    identification, changes in route characteristics and ameliorate burstiness issues
    using buffer-awareness ought to be studied

To top